River level monitoring based on multi-mission altimetry and spatio-temporal kriging – a case study in the Mekong river basin

Eva Boergens (Deutsches Geodätisches Forschungsinstitut der Technischen Universität München (DGFI-TUM), Germany)

CoAuthors

Sven Buhl (Center for Mathematical Sciences, Technische Universität München, Germany); Denise Dettmering (Deutsches Geodätisches Forschungsinstitut der Technischen Universität München (DGFI-TUM), Germany); Florian Seitz (Deutsches Geodätisches Forschungsinstitut der Technischen Universität München (DGFI-TUM), Germany)

Event: 2016 Ocean Surface Topography Science Team Meeting

Session: Science III: Two decades of continental water's survey from satellite altimetry - From nadir low-resolution mode to SAR altimetry, new perspectives for hydrology

Presentation type: Type Poster

Measurements of water level variations of inland water bodies by satellite altimetry got well established in the last years. Many inland water level time series are assembled from measurements of one pass from one single satellite mission. Only a few multi-mission approaches are able to combine different missions and passes over lakes and reservoirs, which increases the accuracy and temporal resolution of the time series. This is possible assuming a constant surface level per epoch. However, it is still challenging to combine different altimeter missions and passes over rivers.

We combine multi-mission altimetry data over the Mekong River using the geostatistical prediction method of ordinary spatio-temporal kriging based on covariances between the altimetry observations. We develop two different covariance models and evaluate their suitability for the Mekong River: a stationary and a non-stationary covariance model.
We use both models for combining altimetry measurements of Envisat, Jason-2, and SARAL. This way we achieve an improved temporal resolution of time series at any given location along the river. The method is validated against in-situ measurements at gauging stations along the Mekong as well as against close by altimeter measurements. Both covariance models yield satisfying results. The RMS values are for the stationary covariance model between 0.75 and 1.15 m and for the non-stationary between 0.62 and 0.74 m. The coefficients of determination are above 0.90 for both models at all validation sites. Furthermore, we show that spatio-temporal kriging is suitable to predict water states for periods with only sparse data coverage, i.e. in the data gap between the end of the Envisat in 2011 and the launch of the SARAL mission in 2013.
 

Poster show times:

RoomStart DateEnd Date
Grande Halle Thu, Nov 03 2016,11:00 Thu, Nov 03 2016,18:00
Eva Boergens
Deutsches Geodätisches Forschungsinstitut der Technischen Universität München (DGFI-TUM)
Germany
eva.boergens@tum.de